Embodiments of the present invention relate to data processing and, more specifically, to selectively processing data associated with a workload.
In some cases, a computer architecture can include multiple clusters of processor cores for processing data. Each cluster can include one or more processor chips, and each chip can include one or more cores. A workload being processed on the computer architecture may thus be processed by one of these cores.
The computer architecture can also include multiple memory storage devices, each of which can be located closer to some clusters than to others. Because communication is required between a cluster and memory being accessed when processing a workload within that cluster, the allocation of that memory can affect performance. Specifically, the speed of processing data associated with a workload may depend on the location of that data in relation to the location of the workload.